SIDESCAN SONAR IMAGE-PROCESSING TECHNIQUES

A four-step processing sequence is described to produce image mosaics from the various segments of a sidescanned acoustic imaging survey of a given seafloor area. Starting with data consisting for each ping of acoustic backscatter levels versus horizontal range across-track, median prefiltering is used first to reduce the influence of outliers on subsequent linear processes. Artifacts that are clearly unrelated to the backscattering properties of the seafloor are then isolated on a ping by ping basis through a spectral analysis that relies on a decomposition using Chebyshev polynomials to filter the low spatial frequency components of the image. Contrast enhancement is then achieved through an original implementation of the classical gray level histogram equalization technique by balancing local versus global histogram contributions. Finally, pixels are mapped on a geographic grid taking due account of the geometry of the measurement and of the spacing between pings to minimize along-track smearing of features. Examples of results obtained with these processing techniques are given for SeaMARC II data recorded during a complete survey of Fieberling Guyot (32-degrees.5N, 128-degrees-W).